Re: TARGET_FPU for mpc8315e-rdb listed as SPE??

[Bruce Ashfield <bruce.ashfield@windriver.com>]

On 10-11-10 09:11 AM, Kumar Gala wrote:
>
> On Nov 10, 2010, at 7:58 AM, Bruce Ashfield wrote:
>
>> On 10-11-10 08:38 AM, Bruce Ashfield wrote:
>>> On 10-11-10 01:46 AM, Kumar Gala wrote:
>>>> Why does the meta/conf/machine/mpc8315e-rdb.conf list TARGET_FPU as
>>>> SPE. This isn't correct for an MPC8313 SoC.
>>>
>>> It isn't used at the moment, so we can safely
>>> ignore this.
>>>
>>> It was a hold over from when I initially created
>>> the BSP, and I've since changed it locally, but
>>> haven't sent the updated BSP yet.
>>
>> To clarify on this point, the kernel configuration
>> is NOT using SPE for this, and I was attempting to
>> use the FPU setting to trigger some different gcc
>> flags during development the base of that test was
>> an e500 board, so the SPE setting leaked in, but is
>> unused.
>>
>> At the moment, it is actually using soft-float, and
>> I had planned to submit a change to clarify that.
>>
>> If there's another option, let me know and I'll
>> rebase my patches and change it again.
>
> We should NOT be using soft-float for mpc8315e (it has HW floating point on this chip).

Indeed. I fell back to a safe multilib. I'm willing
to try again and see if the gcc bootstrap phases will
build.

The kernel was fine, and is fine, it was userspace
that caused problems for me .. and that's definitely not
an area where my expertise lies :)

I needed something that worked, and had to excplicitly
chose to ignore the FPU temporarily, but revisiting that
now seems like a good idea.

>
> Any plans to get an e500 based system going before the rev1.0 release?

We've got tonnes of experience and BSPs to draw from
here. The selection of some of these BSPs was (largely)
based on cost and availability. If we can get our
hands on a suitable e500 replacement .. the switch is
trivial.

Cheers,

Bruce

>
> - k